Stockfish Testing Queue

Finished - 107 tests

19-01-02 Ala ConfinedBishop2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 6739 W: 1389 L: 1520 D: 3830
sprt @ 10+0.1 th 1 Take 2
19-01-02 Ala QueeningSac7 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 30080 W: 6551 L: 6567 D: 16962
sprt @ 10+0.1 th 1 Take 7 : 25, 50 seems to do better than smaller bonuses, try out 40, 60
19-01-02 Ala QueeningSac6 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 19931 W: 4356 L: 4422 D: 11153
sprt @ 10+0.1 th 1 Take 6 : more efficient code + bonus (20, 30)
19-01-02 Ala QueeningSac4 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 32356 W: 6996 L: 7002 D: 18358
sprt @ 10+0.1 th 1 Narrower defended pawn condition
19-01-02 Ala QueeningSac2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 20356 W: 4387 L: 4451 D: 11518
sprt @ 10+0.1 th 1 Take 2 (bonus : 15, 30 instead of 25, 50)
19-01-02 Ala QueeningSac5 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 15046 W: 3253 L: 3343 D: 8450
sprt @ 10+0.1 th 1 Take 5 : narrower conditions + 25, 35 bonus
19-01-02 Ala QueeningSac3 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 10082 W: 2116 L: 2230 D: 5736
sprt @ 10+0.1 th 1 Take 3 (bonus 12, 20)
19-01-01 Ala EgImbalanceTuneB diff
47750/50000 iterations
99567/100000 games played
100000 @ 30+0.3 th 1 Rebased tune. Use latest tune values, add by hand a redundancy penalty for queens and compensate in mg/eg base value. Also change tune range for some values close to 0 to see if they may change.
19-01-01 Ala PPthreats diff
LLR: -2.95 (-2.94,2.94) [0.00,3.50]
Total: 17733 W: 2818 L: 2944 D: 11971
sprt @ 60+0.6 th 1 LTC for Rocky
18-12-31 Ala EgImbalanceTune diff
36088/100000 iterations
75119/200000 games played
200000 @ 30+0.3 th 1 With Rocky's fix
19-01-01 Ala TunedKS diff
LLR: -1.26 (-2.94,2.94) [0.00,4.00]
Total: 25926 W: 4236 L: 4220 D: 17470
sprt @ 60+0.6 th 1 The tuned values failed STC, but the tuning used up the equivalent of 100K LTC games and exhibited a noticeable convergence to a different equilibrium. So test how it behaves at LTC.
19-01-01 Ala PassedPawnTempo2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 17896 W: 3921 L: 3997 D: 9978
sprt @ 10+0.1 th 1 Take 2
19-01-01 Ala PassedPawnTempo3 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 14811 W: 3191 L: 3282 D: 8338
sprt @ 10+0.1 th 1 Take 3 (fixed bench)
19-01-01 Ala PassedPawnTempo diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 13504 W: 2857 L: 2955 D: 7692
sprt @ 10+0.1 th 1 Take 1, bonus when we have a passed pawn more advanced than any enemy passed pawn.
19-01-01 Ala DeadBishop diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 4062 W: 826 L: 970 D: 2266
sprt @ 10+0.1 th 1 Take 1 (fixed)
18-12-31 Ala EgImbalance diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 13964 W: 2986 L: 3082 D: 7896
sprt @ 10+0.1 th 1 Test 30K tune results. Early, but there has already been some significant changes.
18-12-31 Ala EgImbalanceTune diff
12452/100000 iterations
25976/200000 games played
200000 @ 30+0.3 th 1 Try and see if using different imbalances for middlegame and endgame can help
18-12-30 Ala BorderMobilityTuneK diff
96416/100000 iterations
200000/200000 games played
200000 @ 20+0.2 th 1 The bishop values compensate most of the fixed perf hit, tune to see if this can get positive with the knight values.
18-12-31 Ala BorderMobility2 diff
LLR: -2.94 (-2.94,2.94) [0.50,4.50]
Total: 10509 W: 2239 L: 2351 D: 5919
sprt @ 10+0.1 th 1 Test the results of the tune
18-12-27 Ala PassedPawnsArrayB diff
ELO: 1.01 +-2.8 (95%) LOS: 76.3%
Total: 19916 W: 3310 L: 3252 D: 13354
20000 @ 60+0.6 th 1 STC test about to fail. Fixed games LTC to gauge the tune status/progress and inform further action.
18-12-30 Ala BorderMobility1 diff
LLR: -2.96 (-2.94,2.94) [0.00,3.50]
Total: 22959 W: 3733 L: 3845 D: 15381
sprt @ 60+0.6 th 1 STC failed but did +2 elo compared to the perf-loss test. It already used tuning resources, so give it a try at LTC to see if it scales and deserves further work.
18-12-30 Ala BorderMobility1 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 19467 W: 4217 L: 4285 D: 10965
sprt @ 10+0.1 th 1 Test the result of the bishop border mobility tune
18-12-30 Ala ShelterStrength2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 10048 W: 2141 L: 2256 D: 5651
sprt @ 10+0.1 th 1 Test the tuned values
18-12-26 Ala PassedPawnsArray diff
72307/75000 iterations
150000/150000 games played
150000 @ 30+0.3 th 1 Tune based on Bryan's idea of a 2D passed array and on separating passers and candidate passers
18-12-29 Ala BorderMobilityPerfTest diff
ELO: -3.00 +-2.2 (95%) LOS: 0.4%
Total: 40000 W: 8544 L: 8889 D: 22567
40000 @ 10+0.1 th 1 Restart perf-test for new tuning attempt after fixing a bug which plagued the previous try.
18-12-28 Ala ShelterStrength diff
38605/40000 iterations
80000/80000 games played
80000 @ 30+0.3 th 1 Tune rebased on latest master and using latest tune values as a new base.
18-12-29 Ala BorderMobilityTuneB diff
72122/75000 iterations
150000/150000 games played
150000 @ 20+0.2 th 1 Tuning run to see if differentiated mobility bonus for border squares can bring elo (fixed bench).
18-12-29 Ala BorderMobilityPerfTest diff
ELO: -149.12 +-41.0 (95%) LOS: 0.0%
Total: 215 W: 32 L: 119 D: 64
40000 @ 10+0.1 th 1 The potential improvement from the border mobility tune is bigger with more piece type using it, but has a fixed perf hit cost. Evaluate this perf hit cost to better assess the result of the bishop tuning. Low throughput.
18-12-29 Ala BorderMobilityTuneB diff
124/75000 iterations
258/150000 games played
150000 @ 20+0.2 th 1 Tuning run to see if differentiated mobility bonus for border squares can bring elo.
18-12-29 Ala doubledPawnsBlock diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 16773 W: 3665 L: 3747 D: 9361
sprt @ 10+0.1 th 1 Mobility malus for bishops when we have doubled pawns touching each other.
18-12-29 Ala BlockingPawns diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 48162 W: 10586 L: 10513 D: 27063
sprt @ 10+0.1 th 1 Another try to account for enemy pawns blocking bishop
18-12-29 Ala BlockingPawns4 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 7898 W: 1653 L: 1778 D: 4467
sprt @ 10+0.1 th 1 Take 4 : tweaked parameters
18-12-29 Ala BlockingPawns3 diff
6400/25000 iterations
13349/50000 games played
50000 @ 20+0.2 th 1 STC not doing bad with randomly guessed values so try a small tuning run.
18-12-29 Ala BlockingPawns2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 7462 W: 1624 L: 1752 D: 4086
sprt @ 10+0.1 th 1 Take 2
18-12-29 Ala QueeningSquareBishop6 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 10007 W: 2085 L: 2200 D: 5722
sprt @ 10+0.1 th 1 Take 6 : try to make the bonus more balanced to reduce delusions
18-12-28 Ala QueeningSquareBishop diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 28564 W: 6174 L: 6198 D: 16192
sprt @ 10+0.1 th 1 Bonus for a bishop of the passed pawn queening square color when the enemy don't have one.
18-12-28 Ala QueeningSquareBishop3 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 26253 W: 5691 L: 5726 D: 14836
sprt @ 10+0.1 th 1 Take 3 - more mg and less eg bonus
18-12-29 Ala QueeningSquareBishop5 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 20360 W: 4412 L: 4476 D: 11472
sprt @ 10+0.1 th 1 Take 5 : bonus linear to pawn rank rather than quadratic
18-12-28 Ala QueeningSquareBishop4 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 18622 W: 4013 L: 4086 D: 10523
sprt @ 10+0.1 th 1 Take 4 : only bonus for matching bishop, no malus from enemy bishop
18-12-28 Ala QueeningSquareBishop2 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 12628 W: 2699 L: 2801 D: 7128
sprt @ 10+0.1 th 1 Take 2
18-12-27 Ala ShelterStrength diff
24884/50000 iterations
51621/100000 games played
100000 @ 30+0.3 th 1 Try and tune the ShelterStrength array (fixed)
18-12-27 Ala ShelterStrength diff
944/50000 iterations
1988/100000 games played
100000 @ 30+0.3 th 1 Try and tune the ShelterStrength array
18-12-27 Ala PassedPawnsArrayB diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 17069 W: 3659 L: 3739 D: 9671
sprt @ 10+0.1 th 1 Test the values of the tune after 103K games. Random walk for lesser values looks significant at this point, but how does it fare overall ?
18-12-26 Ala BlockedBishop4 diff
LLR: -2.95 (-2.94,2.94) [0.50,4.50]
Total: 2155 W: 437 L: 594 D: 1124
sprt @ 10+0.1 th 1 Take 4 : try a quadratic penalty
18-12-25 Ala BlockedBishop diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 21526 W: 4764 L: 4822 D: 11940
sprt @ 10+0.1 th 1 Count protected enemy pawns on the correct square color as bishop-blocking.
18-12-25 Ala BlockedBishop3 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 11302 W: 2460 L: 2569 D: 6273
sprt @ 10+0.1 th 1 Take 3 : make the center blocked an additive penalty instead of a multiplicative one
18-12-25 Ala BlockedBishop2 diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7776 W: 1710 L: 1837 D: 4229
sprt @ 10+0.1 th 1 Take 2 : bigger penalty
18-12-25 Ala RookMobTest diff
LLR: -2.96 (-2.94,2.94) [0.50,4.50]
Total: 7048 W: 1485 L: 1615 D: 3948
sprt @ 10+0.1 th 1 Remove mobility bonus for rooks when there are restrained by a locked pawn structure. Likely too specific/heavy to pass in its current state, but may give ideas.
18-12-19 Ala Knight8x8b diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 39685 W: 6584 L: 6549 D: 26552
sprt @ 60+0.6 th 1 Test the tuned 8x8 knight PSQT. Don't wait for the STC result to start the LTC test as the tuning was done at LTC.
18-12-19 Ala Knight8x8b diff
LLR: -2.95 (-2.94,2.94) [0.00,5.00]
Total: 10010 W: 2160 L: 2242 D: 5608
sprt @ 10+0.1 th 1 Test the tuned 8x8 knight PSQT.